Currently the study on end-to-end available bandwidth prediction is rare,in light of the status,we propose an algorithm of available bandwidth online prediction (ABOP),which is based on the kernel principle component analysis (KPCA)and the least squares support vector machines (LSSVM).First,the sample data about the network status are collected and reconstructed in phase space.Then,on this basis the KPCA is adopted for data dimensionality reduction and denoising processing.Finally,the online prediction is done on the available bandwidth based on LSSVM.To reduce the computational overhead,we introduce a recursive calculation method to accelerate model update speed,and use particle swarm optimisation to update in multi-step the model parameters to ensure the timeliness of the available bandwidth prediction.Simulations show that the ABOP algorithm proposed can predict in high accuracy and prediction speed,and can meet the requirements of available bandwidth online prediction as well.%针对目前端到端可用带宽预测方面研究工作较少的现状，提出一种基于核主成分分析KPCA（Kernel Principle Component Analysis）和最小二乘支持向量机LSSVM（Least Squares Support Vector Machine）的可用带宽在线预测算法ABOP。在采集网络状态样本数据并对其进行相空间重构的基础上，采用KPCA对数据进行降维降噪处理，最后基于LSSVM对可用带宽进行在线预测。为减小计算开销，提出一种递推计算的方法加快模型更新速度，并采用粒子群优化算法对模型参数进行多步更新，确保了在线预测的时效性。仿真表明，提出的ABOP算法具有较高的预测精度和较快的预测速度，能够满足可用带宽在线预测的要求。
展开▼
